KMS_project

说明：  位图分析转换软件 可以分析8，16，24，32位BMP图 输出可选16位，18位，24位RGB格式 同时支持串行下载(Bitmap analysis of 8,16,24,32-bit conversion software can map out the optional 16-bit BMP, 18 bit, 24-bit RGB format also supports serial download)

使用Visual C++编写的生成三次Hermite曲线的源代码。(Bicubic Hermite code in Visual C++.)

1.打开左右两张影像2.Harris算法提取同名点3.显示提取同名点（若要进行下一步匹配则不建议显示，这里只是用于查看是否提取出）4.相关系数匹配（只在找出的同名点内匹配）5.LSM最小二乘匹配（先进行粗匹配，再精匹配）(1 Open the left and right images 2.Harris algorithm to extract two tie points 3 shows the extraction of corresponding points (not recommended to proceed to the next match show, here is used to see if extracted) 4. Correlation coefficient matching (just looking match within a point of the same name) 5.LSM least squares matching (rough first match, then fine match))

说明：  最二小乘影像匹配，精度可以达到0.01个像素，也可以最二小乘影像模板匹配。经过修改后可以直接用，只要调用一个函数就搞定，函数参数的说明很清楚，绝对不会有任何困难，非常好用。(Most two Theravada image matching accuracy can reach 0.01 pixel images can also be the most two Theravada template matching. After modifications can be directly used, as long as is you call a function, function parameters, the description is clear and will not have any difficulties, very easy to use.)

说明：  很好用的三维重建软件，VisualSFM软件允许上传一系列图像，它从这些图像中找到每一个图像的特定特征，利用这些特征信息重建出3D模型的稀疏点云，而后还可进行稠密点云重建。(The VisualSFM software, a good 3D reconstruction software, allows uploading a series of images. It finds the specific features of each image from these images, and USES these feature information to reconstruct the sparse point cloud of the 3D model, and then it can reconstruct the dense point cloud.)

VC++图像程序处理 是由杨淑莹老师编写的 对于计算机视觉 等课程有重要作用 (VC++ image processing program written by Yang Shuying teacher courses for computer vision plays an important role)

利用直方图双峰法选取适当阈值对图像进行二值化处理。此法只适用于直方图为双峰状图像。(Using the Shuangfeng method to the appropriate threshold histogram of image processing two values. This method is only applicable to the histogram of image of Shuangfeng.)

该函数用来实现快速离散余弦变换。该函数利用2N点的快速付立叶变换来实现离散余弦变换(This function is used to implement fast discrete cosine transform. 2N points of the function using the fast Fourier transform to achieve the discrete cosine transform)

先进行图像预处理，比如Hough变换，边沿检测，还有差影法，模板提取 ，识别，跟踪，最后是模板匹配。(First image preprocessing, such as the Hough transform, edge detection, as well as poor shadow method, template extraction, identification, tracking, and finally the template matching.)

特征点匹配时，采用构造KD-tree，用最优节点优先法寻找最佳匹配点(Matching feature points, the tectonic KD-tree, with the optimal node is preferred method to search for the optimal matching point)